Research map / Image generation / Text-to-image diffusion
MaxFusion: Plug&Play Multi-Modal Generation in Text-to-Image Diffusion Models
Nithin Gopalakrishnan Nair, Jeya Maria Jose Valanarasu, Vishal Patel
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fuse control features by variance-based selection instead of naive averaging across conditions | Stable Diffusion 1.5 with ControlNet or T2I-Adapter · COCO contradictory conditions (Pose, Seg) | FID: 55.61 -> 46.72 (vs T2I-Adapter) | Improved |
| use MaxFusion on Pose + Depth contradictory conditions | Stable Diffusion 1.5 · COCO contradictory conditions (Pose, Depth) | depth MSE: FID 48.18 -> 47.37, MSE-D 0.0357 -> 0.0410 (vs ControlNet) | Mixed |
| MaxFusion instead of T2I-Adapter averaging for complementary conditions | Seg + Depth to image · COCO | Depth MSE: 0.0554 -> 0.0296 | Improved |
| MaxFusion instead of T2I-Adapter averaging for HED + Depth | HED + Depth to image · COCO | CLIP score: 0.2964 -> 0.2879 | Hurt |
| MaxFusion instead of T2I-Adapter averaging for HED + Seg | HED + Seg to image · COCO | FID: 40.14 -> 37.21 | Improved |
| extend MaxFusion incrementally to three modalities | Stable Diffusion 1.5 · COCO (Seg, Depth, HED) | FID: 50.47 -> 47.48 (vs ControlNet) | Improved |
| correlation threshold set to 0 (naive averaging) versus higher thresholds (ablation) | MaxFusion · qualitative teddy bear example | depth and edge conditioning strength | Improved |
| combine edge and depth conditions rather than using a single condition (ablation) | MaxFusion · qualitative bird example | image quality | Improved |
| naive feature averaging for complementary conditions (ablation) | multi-condition diffusion · qualitative complementary conditions | image quality | No clear effect |
Later papers that compared with this method
1 later accepted paper reports MaxFusion as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Cross-ControlNet: Training-Free Fusion of Multiple Conditions for Text-to-Image Generation (ICLR 2026 · better on 7 of 8 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- T2I-Adapter — T2I-Adapter: Learning Adapters to Dig Out More Controllable Ability for Text-to-Image Diffusion Models (AAAI 2024) · better on 9 of 11 shared results
- ControlNet — ControlNet++: Improving Conditional Controls with Efficient Consistency Feedback (ECCV 2024) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Dual Diffusion for Unified Image Generation and Understanding (CVPR 2025 · similarity 0.91)
- Image Regeneration: Evaluating Text-to-Image Model via Generating Identical Image with Multimodal Large Language Models (AAAI 2025 · similarity 0.91)
- Exploring the Deep Fusion of Large Language Models and Diffusion Transformers for Text-to-Image Synthesis (CVPR 2025 · similarity 0.90)
- R&B: Region and Boundary Aware Zero-shot Grounded Text-to-image Generation (ICLR 2024 · similarity 0.90)
- One-Way Ticket: Time-Independent Unified Encoder for Distilling Text-to-Image Diffusion Models (CVPR 2025 · similarity 0.90)
- T2I-Adapter: Learning Adapters to Dig Out More Controllable Ability for Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
